3D rig for HDR cinematography?

[Rudi Herbert]

Would it be possible to use a 3D rig to shoot HDR images instead? I'm not interested in shooting 3D but instead, there's a feature we'll start shooting later in the year where 90% of the script takes place outdoors, and we really want to extol the beauty of the location, so we want crisp, blue skies and detailed highlights together with well exposed, rich shadow details on the costumes and skin tones.

In digital cinematography, you either expose to preserve highlights or shadows, but with two cameras shooting the same image, each of them exposed to favor one end of the spectrum, wouldn't we in theory get the equivalent of HDR cinematography? maybe this rig can be set up so that both lenses can be operated by the same follow focus, maybe by reversing the gear rotation on one of the lenses or something to that effect, so that the same operator can pull focus on both cameras at once. Yes, doubling the weight of the camera and lens would make it a bulky, heavy contraption, and double your expenses in camera bodies and optics as well as doubling the amount of data you have to deal with in post. But, such a rig would still be manageable and the increase in expense and data wrangling would be more than justified if you end up with an image that gives you that impossibly extended dynamic range, potentially something in the 14-16 stops range, and provide a result that simply cannot be emulated in post. Opinions?

 

[Dan Hudgins]

Using two RED ONE for HDR...

Using two RED ONE for HDR...

I started a long thread some time ago about doing this.

http://www.reduser.net/forum/showthread.php?t=14479

http://www.reduser.net/forum/showthread.php?t=14479&page=3

http://www.reduser.net/forum/showthread.php?t=14479&page=4

http://www.reduser.net/forum/showthread.php?t=14479&page=5

http://www.reduser.net/forum/showthread.php?t=14479&page=7

http://www.reduser.net/forum/showthread.php?p=234479#post234479

You need to look at all the pages in that thread and some others I have made on the subject, you can use google on a search of REDUSER and HDR etc.

My programs can FUSE the two images after the hi and low frames are graded. You could probably do that in other programs as well. To do it with mine you would grade to TIF output then FUSE at tri-color frames, the re-grade the fused frames for S curve and chroma matrix etc. to get good results in the final HDR frames. In my program you can adjust the line up of the frames manually if you crop a little and enlarge enough to make up the frame size.

I posted some drawings of the beam splitter options, you would want a beam splitter that does not degrade the images or make artifacts, which I went into. There is also the issue of how best to use wide angle lenses which I went into also. If you want to talk about details you can email me. See the link to page 5 above and sign in to see the drawings "45*.jpg" etc...

 

[David Birdy]

Most 3D rigs are built to have two cameras, len's, focus and iris all work together in tandem.

Dave

 

[Dan Hudgins]

Solarization?

Solarization?

I noticed in one shot in "Che (2008)" the car headlights looked like they had Solarization, this is the spot that should have been blown white was dark, has anyone else got the "dark sun" effect in their RED ONE?

If the current builds of the firmware still let some Solarization happen that could be an issue with doing HDR with the camera.

Older CMOS sensors had some "dark sun" issues, other than that one shot I have not seen the problem in RED ONE footage, and the headlights must have been more than 3 stops over white clip level, maybe?

 

[Stefan Scherperel]

A 3D rig would not work for HDR cinematography becasue the cameras are not capturing the image from the same exact location. For subjects that are more than about 30ft away it may not be an issue, but subjects that are closer to the camera will be captured from to different angles and thus not line up for combinng together for HDR.

 

[Dan Hudgins]

Beam splitter 3D rig...

Beam splitter 3D rig...

A 3D rig would not work for HDR cinematography becasue the cameras are not capturing the image from the same exact location. For subjects that are more than about 30ft away it may not be an issue, but subjects that are closer to the camera will be captured from to different angles and thus not line up for combinng together for HDR.

At least from what I was talking about was a beam splitter 3D rig like was used to shoot 3D movies with movie cameras, they let the cameras eye spacing go down to zero if built right (you need to be able to go down to zero spacing for close up work and macro 3D shots anyway).

So what kind of rig are you thinking about?

Anyway if you look at the drawings in the links above you will understand what type of rig would work...

 

[Rudi Herbert]

Dan,

I read through all the posts in the thread you forwarded, thanks. I have to clarify that, although I much appreciate the contributions of such a scholarly fellow like yourself, my interests are perhaps a bit more practical and, more importantly, urgent. I don't care for arrays with 3, 4, 6 or 9 RED cameras, however, I am sure that something like that could potentially yield amazing results. I am nevertheless VERY interested in a set up with two cameras, for this is something I wish to consider for use just a few months from now. So, much more than "what could be accomplished" I am interested in "what can be done", for I believe that HDR cinematography, or a derivative of it in some shape or form, is the only way for digital pictures to surpass film based acquisition. With that in mind:

How about a rig where we have just one lens, hooked to a PL mount right in the center of a device that has a mirror/splitter that feeds exactly the same image to two cameras set up side by side? Maybe one side of this mirror/splitter lightens or darkens one half of the image fed to one of the cameras and that corresponding camera has its shutter angle set up so that it captures an image with favors the opposite side of the spectrum as the other camera? I feel that, if we can use one lens and a contraption that ties up two camera bodies side by side, where the only increase in weight is one more body, and in expense, then this becomes quite a feasible proposition for most productions. At least it would be for me. I've been quietly testing cameras for the past few months, based on the premise that, when it's time to go, I would prefer to go with RED but I'll take whatever does the job best provided I can afford it. I've been blown away with F35 footage that I received and I've been playing with, that camera is still MILES ahead of anything out there, including RED, but it is still about 2-3 1/2 stops better than RED, not more, and that to me, is still way inferior than what could be attained with an HDR rig. So, again, if anybody can produce a rig that successfully uses two RED bodies, preferably using only one lens, than can yield such images, please contact me because you got yourself a customer.

 

[Dan Hudgins]

f/ratio

f/ratio

I mentioned in one of my posts before I think making an adapter that has two PL mounts going into two RED ONE bodies, and one PL mount for a lens of any focal length.

I am not a job shop, but I may be able to help you with the drawings and I know a machinist that makes/made things for ILM that might be able to produce the units, his rates may not be too high. If the drawings are good enough almost any machine shop can make the parts if you have room for shims to adjust the placement of the optics.

Before you get too worked up over the idea, you should test the over exposure range of the RED ONE, as if you get the "dark sun" effect that will be an upper limit, see if you can get 10 stops over white clip for the "high" camera.

I can try processing TIF frames from the high and low exposures for you to see what the results will look like, we can make the exposure spread for the tests in half stop gaps, like +/-1 stop, +/-2 stops, +/-3 stops and so on to +/- 6 stops. +/-6 stops is about the limit since you need 6 bits to get 64 tones which is the minimum for an image.

To make a "T" shaped adapter you need two prime relay lenses focused at INF on each RED ONE camera, Nikkon or Schneider enlarger lenses could work as they should be sharp, the only issue is that would limit the max iris to about 2.8. You could use f/1.2 Canon lenses as the two prime lenses but you would get some fall off. You then go through a 90/10% or 95/5% cube Beam splitter and then another large opening lens, then a field lens, the field lens would need to be swapable to alow for both wide and long lenses to keep the corners of the frame sharp, and then the PL mount or Nikkon/Canon mount and whatever lens you want of any focal length.

Being machined out of solid metal will help keep the two cameras close to a few pixel in alignment and you can make up the small errors in software. It would be good to avoid having to ratate in software, but you will probably will, so there will be some small loss of resolution, but the RED ONE frames are not single pixel sharp, so I don't thing it will be a show stopper.

All the optics requires should be able to be taken from stock parts, the metal comes from a metal yard, and the work can be done in more than one shop for the best bid. You could make more than one and sell the extras to recover the cost of making one for yourself so it could cost you nothing, maybe.

Like I said, the only problem I see is the max f/stop impacting the cost and size of the unit, if you can get by with f/4.5 or something it would be smaller than trying to get f/1.2...

 

[J. Eric Camp]

I have been working up designs for a while on this. Two bodies are good. Three bodies are great.
It all needs to happen behind the pl mount in order to make the rig manageable and elegant. Thats a problem. Then it's in front of the lens and using multiple lenses.

 

[Dan Hudgins]

parrallel

parrallel

I have been working up designs for a while on this. Two bodies are good. Three bodies are great.
It all needs to happen behind the pl mount in order to make the rig manageable and elegant. Thats a problem. Then it's in front of the lens and using multiple lenses.

the rays through the cube beam splitter must be parallel, so you need three imaging lenses in addition to the primary PL/Nikon/Canon lens and the field lens.

The two lenses on the camera bodies are matched and should be about 85mm to 250mm, the 3rd lens would the the same focal length but should have a larger opening to avoid vignetting. Three f/1.2 85mm lenses might work if you limit the two on the cameras to f/2.8 or maybe f/2.0. You can run everything wide open, but there will be some extra falloff since the camera lenses are spaced from the reverse mounted lens on the other side of the cube prism.

Using a cube prism reduces the dust problems and vibration problems with a Pellicle beam splitter.

There are no "problems" with building one of these rigs, just some limits on how wide you can open the iris with stock optics. If you have money, and want to design an all new optical system, you can open up the optics and get maybe f/1.2 through, although the T loss is going to be about a stop no mater what the f/stop is.

 

[J. Eric Camp]

I was leaning towards the Pellicle if it wasn't for the vibration. Dust can be dealt with with proper manufacturing.

When I say problem, I mean more R/D and or money for manufacturing.

 

[Dan Hudgins]

cost...

cost...

I was leaning towards the Pellicle if it wasn't for the vibration. Dust can be dealt with with proper manufacturing.

When I say problem, I mean more R/D and or money for manufacturing.

You can buy a shoptask for like less than $2000 maybe less used, the metal might cost $500 at a surplus lot. What R/D?

 

[J. Eric Camp]

All based on my situation. I live in an apartment. Can't buy a shoptask and put it in the bedroom. Also didn't figure I'd get things like the FFD right the first time around if I was making a block behind the lens. Maybe I'm an idiot though.

 

[Dan Hudgins]

Ffd

Ffd

All based on my situation. I live in an apartment. Can't buy a shoptask and put it in the bedroom. Also didn't figure I'd get things like the FFD right the first time around if I was making a block behind the lens. Maybe I'm an idiot though.

If you use two SLR lenses on the REDONE's then you do not need to deal with the FFD there, you can attach the block to the front of the lenses that are set to INF with a filter ring that you clamp on with set screws from outside so you just need to make two rings, those can screw to the block holding the prism. On the other end you can adjust the FFD by focusing the thrid SLR lens a little and using shim stock behind the PL mount and by moving the field lens in and out a little.

People rent garages. You could make friends with some one who has a machine shop and rent time there...